Claudia Zanella^{1}, Christian T Stoeck^{1}, Constantin von Deuster^{1}, and Sebastian Kozerke^{1}

Higher order diffusion imaging has revealed new insights into myocardial microstructure. At the b-values required for diffusion kurtosis or diffusion spectrum imaging, translation into human in-vivo application requires Stimulated Echo Acquisition Mode imaging (STEAM) at prolonged mixing-times. In this study the effect of mixing time on diffusion parameters is investigated covering the range of current spin-echo and STEAM aproaches. Results show that fractional anisotropy increases and mean diffusivity decreases with mixing-time. Diffusional kurtosis was found to decrease with mixing-time by varying amounts along fiber, sheet and sheet-normal direction which needs to be considered when comparing spin-echo imaging with STEAM acquisitions.

**Methods**

Data acquisition

One porcine heart was perfusion- and immersion-fixed with a 4% PFA2 in phosphate-buffered saline solution within 20min after excision. Imaging was performed on a 3T clinical system (Philips Healthcare, Best, The Netherlands) equipped with an 8-channel head coil and a gradient system delivering 80mT/m @ 100mT/m/ms. DSI data was acquired using a single shot echo planar imaging (EPI) STEAM sequences with a half sphere sampling pattern: bmax=10015s/mm2 and an edge length of 15 along each dimension. Imaging parameters were as follows: 2 slices, resolution: 2.5×2.5×8mm3, TR: 2s, TE: 81ms. An unweighted reference image was interleaved every 20th acquisition and the order of diffusion encoding was optimized to minimize gradient duty cycle4. Timings of the unipolar diffusion encoding gradients are reported in Table 1. Complex averaging was employed to avoid bias due to noise floor5. Over the course of the entire experiment temperature was monitored via optical fiber thermometry inside the sample at 10sec intervals.

Data analysis

Diffusion tensors were estimated upon manual masking of the myocardium including data with a bmax<2000s/mm2. Pixels for which the tensors first eigenvector deviated from the circumferential course by more than 30° were discarded. Mean Diffusivity (MD), Fractional Anisotropy (FA) as well as the tensors’ eigenvalues and eigenvectors (v1/v2/v3) were derived from the diffusion tensors. Displacement probability distribution functions (PDF) along v1/v2/v3, i.e. fiber, sheet and sheet normal direction, were computed and the Full Width at Half Maximum (FWHM) calculated upon Hamming filtering and zero filling (1.5 times) of q-space. Diffusional kurtosis K was calculated by fitting: $$log\left(\frac{S\left(b\right)}{S\left(0\right)}\right)=-bD+\frac{Kb^2D^2}{6}$$ to the resampled signal along v1,v2 and v3 (bmax<5000s/mm2)3. S(0) denotes the unweighted reference image.

**Discussion**

1. Mazzoli V, Froeling M, Nederveen AJ et al. Cardiac diffusion MRI beyond DTI; ISMRM 2014

2. Lohezic M, Teh I, Bollensdorff C et al. Interrogation of living myocardium in multiple static deformation states with diffusion tensor and diffusion spectrum imaging. Prog Biophys Mol Biol. 2014;115(2-3):213-25

3. McClymont D, Teh I, Carruth E et al. Evaluation of Non-Gaussian Diffusion in Cardiac MRI; MRM 2016; doi: 10.1002/mrm.26466

4. Vos SB, Tax CMW, Luijten PR, et al. The importance of correcting for B0-drift-induced global signal decrease in diffusion MRI. MRM 2016; doi: 10.1002/mrm.26124

5. Scott A, Nielles-Vallespin S, Ferreira PF, et al. The effects of noise in cardiac diffusion tensor imaging and the benefits of averaging complex data; NMR in Biomed 2016 ;29(5):588-99

6. Kim S, Chi-Fishman G, Barnett AS et al. Dependence on diffusion time of apparent diffusion tensor of ex vivo calf tongue and heart; MRM 2005; 54(6):1387-96

7. JensenJH, Helpern JA MRI quantification of non-Gaussian water diffusion by kurtosis analysis.; NMR Biomed. 2010;23(7):698-710

8. von Deuster C, Stoeck CT, Genet M, et al. Spin echo versus stimulated echo diffusion tensor imaging of the in vivo human heart. Magn Reson Med. 2015;doi: 10.1002/mrm.25998

Table 1 Gradient timings and resulting maximal q value (corresponding
to bmax of 10015s/mm^{2}). The number of averages was
adjusted to compensate expected signal loss due to T_{1} relaxation during
the mixing time.

Figure 1: Maps of mean diffusivity (MD), fractional anisotropy
(FA) and kurtosis along the principal (K1), secondary (K2) and tertiary (K3)
eigenvector of the diffusion tensor. For illustration purpose, missing pixels
for which the deviation of the first eigenvector from the circumferential
structure deviated by more than 30° have been interpolated from neighboring
pixels.

Figure 2: Mean
Diffusivity (MD), Fractional Anisotropy (FA) and Eigenvalues derived from the
tensor model for different gaps between the diffusion encoding gradients (TM).
The vertical dotted line indicates a change in TM-interval.

Figure 3: Example displacement probability distribution
functions (PDF) along the diffusion tensors‘ eigenvectors averaged over the
myocardium (left). Full width at half maximum (FWHM) of the PDFs as function of
the square root of the effective duration plus the gap between the diffusion
gradients (TM) (right). A linear fit for
TM≤80ms is presented by the dotted line.

Figure 4: Diffusional kurtosis along the direction of the eigenvectors
of the diffusion tensors. The vertical dotted line indicates a change in
TM-interval.